Technology & Research

Intel® Technology Journal Home

Volume 12, Issue 03

Original 45nm Intel® Core™ Microarchitecture


Intel Technology Journal - Featuring Intel's recent research and development

ISSN 1535-864X DOI 10.1535/itj.1203.01

  • Volume 12
  • Issue 03
  • Published November 7, 2008

Original 45nm Intel® Core™ Microarchitecture

  Section 5 of 10  

Original 45nm Intel® Core™2 Processor Performance

INTEL® CORE™ PROCESSOR

Desktop and Mobile

The PenrynΔ family of processors, including dual- and quad-core desktop processors and a dual- and quad-core mobile processor are branded as ‘Intel Core processors.’

Desktop and mobile systems built with 45nm Intel® processors, based on Penryn Core architecture, give gamers, researchers, and serious multitaskers a significant performance boost over previous-generation processors. In this section of the paper we present measured performance data on next-generation Intel® Core™2 Extreme processors, the new addition to Intel's high-end desktop product line-up. We compare this processor with previous-generation Intel Core 2 Extreme processors on key client benchmarks and real-world applications.

Microarchitectural performance

Improvements

(Figure 6) shows a comparison between the Intel Core 2 Extreme QX6850 processor (3.00GHz, 1333MHz FSB, 8MB L2) and the next-generation 45nm Intel Core 2 Extreme QX9650 (3.00GHz, 1333MHz FSB, 12MB L2) on an Asus* P5EX38 board at the same frequency and platform configuration. 45nm quad-core performance is up to 6 percent faster than previous-generation technology on SPEC 2006.



Figure 6: Quad-core performance comparison for SPEC CPU2006 at same frequency (estimated SPEC CPU2006 as measured on pre-production systems).

Video and audio encoding are becoming increasingly important in the world of personal computing. Home-editing of videos and sound recordings are among the popular applications as is standard archiving of DVD material. As shown in (Figure 7) , the 45nm Intel Core 2 Extreme QX9650 provides a significant boost over previous-generation processors at the same frequency and platform configuration for some of the media-encoding applications. For example, Premiere* Pro CS3 software from Adobe is used to create high-quality visual and editorial effects; it allows users to add color correction, lighting, and other effects such as audio filters and more, with fast, flexible, built-in tools. As shown in (Figure 7) , the new Qx9650 is 20 percent faster than the Qx6850 in rendering 210 frames to the disk using this Adobe software. Fathom* is an advanced encoding platform product from Inlet Technologies that is used by media companies to encode content for streaming over the Internet or broadcasting over the air. As shown in (Figure 7) , Intel measures a 23-percent improvement with new 45nm processors for Fathom to transcode 1080i YV12 high-definition video (HDV) to a 1080i VC1 format. Intel measures a 40-percent improvement for Qx9650 over previous-generation technology for a Pegasys* TMPGenc XPress 4.4 encoder to convert original Variable Bit Rate encoded, 76 second, 29.97fps, 1440×1080 video clips into HDV format MPEG video with 1440×1080 resolution, 29.97fps, and 25000Kb s Constant Bit Rate encoding. Another example is VirtualDub* software, which is a video capture processing utility that uses the DivX* 6.7 software for encoding movies. VirtualDub* 1.7.1 and later with DivX 6.7 are optimized for SSE4 instructions and provide a very noticeable 60-percent performance gain over previous-generation processors that use encoding in SSE2 to convert to the higher-compression DivX format.



Figure 7: Quad-core performance comparison for video encoders at same frequency.

(Figure 8) shows a similar comparison for some of the popular games. The new 45nm Intel® Core™2 Quad processor is roughly 10 percent faster than previous-generation processors at 1024×768. Even when looking at just the two quad-core processors that run at the same FSB and clock speeds, the Intel 45nm Core 2 processors have a clear lead over previous-generation processors. The larger cache, new microarchitectural features, the Penryn high-definition boost, the Super Shuffle unit, and the SSE4 instructions that were discussed in the previous sections all contribute to the increased performance.



Figure 8: Quad-core performance comparison for games at same frequency.

Frequency and platform improvements

Figures 9 and 10 compare the Intel Core 2 Extreme QX6850 processor (2.93GHz, 1066MHz FSB, 8MB L2) on an Intel® 975BX2 board with DDR2 800 RAM with an Intel Core 2 Extreme QX9770 processor (3.20GHz, 1600MHz FSB, 12MB L2 Penryn). Please see Table 1 for a detailed system configuration. This is a more realistic comparison as it takes into account core enhancements, frequency improvements achieved with new 45nm technology, and other platform improvements that were added to support core enhancements. A new 45nm Intel Core 2 Quad-based platform provides double-digit gains on compute-intensive workloads such as SPEC 2006 and the Sysmark 07 Preview that reflect usage patterns of business users in the areas of video creation, E-learning, 3D modeling, and office productivity. Intel measures a 17 percent improvement for Cinebench's multi-threaded rendering test and roughly a 20 percent improvement for Quake 4* and Half Life 2*. A video-encoding application such as DivX and TMPGenc* see a 50 percent to 80 percent gain. The increased frequency and 1600-MHz FSB improves the system bus and memory bandwidth and are the significant contributors to the performance difference. The enhancements in the Penryn family of processors are setting milestones in desktop computing performance.



Table 1: Detailed system configuration for the results shown in Figures 9 and 10



Figure 9: QX9770 (45nm) comparison with QX6850 (65nm) or SPEC 2006.



Figure 10: QX9770 comparison with QX6850 for games and general purpose applications.

Enhanced Intel® Dynamic Acceleration Technology

Performance data presented in Figures 6–10 are based on desktop system measurements, but mobile platforms built with 45nm cores will see similar performance improvements over previous-generation platforms. Mobile processors, however, operate at lower frequencies because of tighter power and thermal limitations. In this section we illustrate how 45nm enhancements in Intel® Dynamic Acceleration Technology (DAT) improve mobile platform performance.

DAT is a power-management feature that can improve system performance by increasing the frequency of active core(s) when at least half of the cores in a multi-core processor are inactive and thermal headroom is available. DAT was introduced in the 65-nm Intel Core 2 mobile processors, but in the Penryn family of processors we further enhanced DAT performance and energy efficiency by reducing the number of transitions in and out of DAT, reducing transition overhead in high-interrupt-rate scenarios. In the Penryn family of processors, we also extended DAT support to quad-core mobile processors. Architectural implementations and more details about these enhancements in Intel's Enhanced Dynamic Acceleration Technology (EDAT) are discussed in.[4] Single-threaded applications running on a dual-core or quad-core processor based on the Penryn family of processors, or two single applications (or a two-threaded application) on a quad-core processor, can take advantage of EDAT.

Figures 11-12 illustrate EDAT performance on an Intel Core 2 Quad processor for SPEC 2006, games, and multimedia applications on pre-production mobile platforms with 2GB of DDR3 memory, a 1066MHz FSB, a 120GB hard disk, and a baseline frequency of 2.4GHz.



Figure 11: EDAT performance improvements (estimated SPEC 2006 as measured on pre-production hardware)



Figure 12: EDAT performance improvements (multimedia and games as measured on pre-production hardware)

To fit QC processors into the mobile Thermal Design Power (TDP) envelope, Intel had to reduce the processor's operating frequency, which in turns gives a wider dynamic range in terms of thermal headroom to operate at EDAT frequencies for one- and two-core operations. The performance gains for EDAT are solely from frequency scaling. The amount of frequency improvement varies from product to product. Single-threaded applications, or two single-threaded applications, or an application with two worker threads with good frequency scaling running on a quad-core processor, can see up to a 10 percent performance boost from EDAT.

  Section 5 of 10  

Back to Top

In this article

Download a PDF of this article.